4
$\begingroup$

I am working with gene expression microarrays of tumor tissues and I want to use a program to find the clusters of co-expressed genes in order to know if some particular genes are co-expressed with another genes and which genes are those.

As I must do this for many microarray experiments and I have read that there are a lot of genes that have constant expression in a tissue, I ranked the gene by its variability (coefficient of variation) that they had in the experiments and keep the first 3000 genes to search their co-espression. Then I employed a program (a bioconductor package) to find co-expression.

Now, I was expecting to find several clusters of genes (no particular reason for this) in the experiment that I was using as example, but instead of that, I only find one cluster of co-expressed genes of about one hundred, and the others genes didn't were co-expressed.

My question is: this result could have a biological sense, or I'm making a terrible mistake in some place?, Thanks in advance.

Improve this question
$\endgroup$
3
  • $\begingroup$ You clustered the genes based on what? $\endgroup$
    – WYSIWYG
    Jul 11 '16 at 15:33
  • $\begingroup$ It's really not possible to interpret your analysis when we don't know what the data is. How many samples (array hybridizations) are there? What are the conditions, do you expect a good diversity of expression profiles? Is there any indication that the data is good quality, like concordant replicates? Many publicly available data sets are small / poor quality and cannot be used for cluster analysis. $\endgroup$
    – Roland
    Jul 14 '16 at 7:15
  • $\begingroup$ Also, you might want to check out Olga Troyanskaya's lab at Princeton, reducio.princeton.edu/cm/ogt They have worked on large-scale coexpression problems for some time and provide software libraries that solve many of the basic problems. $\endgroup$
    – Roland
    Jul 14 '16 at 7:16
1
$\begingroup$
  • Check the beautiful publication of Daniel Ramsköld et al. 2009, which holds the numbers for generally anticipated co-expression.
  • The specific level of co-expression, which applies to your scenario, will depend upon your tissue, your thresholds, and your definition of co-expression.
  • It you look for a co-change of some genes across different specimen (rather than co-expression), the number of genes that should change will depend upon the underlying biology (and thus prevent a general answer without considering the specifics of your specimen).
Improve this answer
$\endgroup$
0
$\begingroup$

There are certain very well defined groups of genes that you would expect to be co-expressed, e.g. ribosomal proteins, proteosome subunits, splicesome components, VHATPase subunits; and each of these groups are co-expressed in different circumstances. It therefore looks to me that either your different tumour tissues (if that is the situation) are in similar conditions so you can’t differentiate these clusters, or that there is some problem with your analysis.

Improve this answer
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.